In a lithography process which is a part of a process of generating a semiconductor circuit, a photosensitive light-ring resist (hereinafter, referred to as “resist”) is applied onto a substrate such as a semiconductor wafer, with a predetermined thickness, reduction exposure of a mask pattern is performed by using an exposure device, development thereof is performed, and thereby a resist pattern is generated. A shape of the generated pattern changes due to various process variations, and thus, the pattern is managed by measuring a dimension of the pattern using a scanning electron microscope (hereinafter, referred to as “length measuring SEM”) with a dimension measurement function.
However, recently, a variation allowable amount of the pattern has been reduced in accordance with thinning and densification of a semiconductor device. Hence, simple dimension management is not sufficient, and it is necessary to manage variation of a focus value and an exposure amount of the exposure device which is a part of process variation factors, with high accuracy.
PTL 1 discloses a process window analyzing method of performing measurement or shape evaluation of a pattern size of each exposure condition, and calculating exposure conditions within a variation allowable range of a pattern, using a focus exposure matrix (FEM) wafer whose pattern is printed by changing the exposure conditions (focus, exposure amount) for each shot (exposure unit of one time), as a method of managing variation of the focus value and the exposure amount of the exposure device. By using the analyzing method, variation of the exposure conditions can be appropriately evaluated, and management of manufacturing conditions can be performed. In addition, PTL 2 proposes a method of extracting various feature amounts from a pattern shape of each exposure condition which is acquired by using the FEM wafer, generating an estimation model by using the feature amounts, and estimating exposure conditions of a management target pattern by using the extracted feature amounts extracted from the imaged management target pattern and the estimation model.